How Theory of Mind outperforms associative learning

Krebbers, Marin (2025) How Theory of Mind outperforms associative learning. Bachelor's Thesis, Artificial Intelligence.

Preview	Text bAI2025KrebbersM.pdf Download (2MB) | Preview
	Text Toestemming.pdf Restricted to Registered users only Download (195kB)

Abstract

Theory of Mind (ToM) is the ability to attribute mental states to others. This ability is fundamental to human social cognition, but its presence in animals remains an open question, as observed signs of ToM might instead be explained by associative learning. This study investigates whether ToM reasoning provides a performance advantage over purely associative strategies in unpredictable environments. Using a negotiation game in the Coloured Trails environment, we compare Deep-Q Network (DQN) agents (zero-order ToM) with first-order ToM agents that simulate their opponent’s decisions using an internal DQN model. Despite limited success in accurately inferring opponent goals, ToM agents generally achieve higher cumulative rewards than DQN agents across varied conditions. While the advantage is not consistent across all scenarios, these results indicate that predictive ToM reasoning can provide a meaningful advantage in certain contexts. This supports the idea that ToM-like strategies may have emerged gradually from associative learning, which would make it easier to believe that animals have ToM.

Item Type:	Thesis (Bachelor's Thesis)
Supervisor name:	Weerd, H.A. de
Degree programme:	Artificial Intelligence
Thesis type:	Bachelor's Thesis
Language:	English
Date Deposited:	29 Jul 2025 06:33
Last Modified:	05 Aug 2025 07:35
URI:	https://fse.studenttheses.ub.rug.nl/id/eprint/36483

Actions (login required)

View Item